Code graph tool efficiency varies by AI model and repository size in benchmark

·1 views

A developer benchmarked a code graph tool against standard grep for AI coding agents across four public repositories of varying sizes. The test used Claude's Haiku, Sonnet, and Opus models to answer structural and exact-match coding questions. For large repositories like Kubernetes and VS Code, the Haiku model with the graph tool was significantly cheaper and more accurate on structural queries. However, the Sonnet model saw little benefit on large projects and higher costs on small ones, while Opus showed negligible efficiency gains.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Techrights article claims IBM is gradually phasing out Red Hat

A Techrights article published on October 1, 2026, alleges that IBM is slowly phasing out Red Hat. It draws parallels to other companies IBM has acquired. The article was discussed on Hacker News, where it received nine points and six comments. The original claim presents IBM's acquisition strategy as leading to the elimination of acquired brands. No official statement from IBM or Red Hat is cited in the provided source.

0 comments Read more at Hacker News

ProgrammingDEV Community ·

Developer showcases year of AI tool projects, including RAG and automation apps

Developer Sufiyan from Satara, India, has spent the past year building several AI-integrated applications. His primary project, Groundwork, is a retrieval-augmented generation tool that answers user questions based on uploaded documents. Other tools include LeadAI for automated lead follow-up, AI CodeReview Pro for analyzing code changes, and the open-source NotesToReport. He has made all projects available with live demos and code on his portfolio site and is now accepting freelance work in AI automation and full-stack development.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Chatbot improvement evaluation stalled due to API credit shortage

A developer has been unable to measure improvements to a specialized agricultural chatbot for over two weeks. The bot, designed to answer pasture growth questions for rural workers, initially failed 67 out of 68 test questions in a baseline evaluation. While 22 specific issues were identified and many have been addressed, a follow-up audit was halted due to depleted API credits. Consequently, the developer knows what fixes were made but cannot confirm if the chatbot's performance has actually improved.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

GitLab Enterprise Edition patched critical vulnerability exposing search credentials

GitLab disclosed a critical vulnerability in its Enterprise Edition on September 10, 2026. The flaw, tracked as CVE-2026-87719, involves insecure deserialization and carries a CVSS score of 9.9. An authenticated attacker could exploit it via GraphQL to access Advanced Search configuration and sensitive credentials. GitLab released patches in versions 19.3.2, 19.2.6, and 19.1.8, with backports to earlier releases. The company advises administrators to upgrade immediately and rotate exposed credentials.

0 comments Read more at DEV Community